The reproductive output of plants depends on the fine-tuning of flowering to fit abiotic and biotic conditions [57, 58]. Thus flowering phenology has strong fitness consequences, and flowering time is one factor determining a species' niche . Like other characters specifying a species' niche, physiological or morphological, flowering phenology may be conserved, plastic or undergo evolutionary change [57, 59].
It is well established that many plant populations can and have changed their flowering time in the last century. Shifts in flowering phenology provide some of the most compelling evidence that species are being influenced by contemporary global environmental change [60–62]. These tracking responses, perceived as adaptations to changing environmental conditions, may be environmentally induced plastic responses or evolutionary adaptations. Although genetic data is not available for most of the species showing phenological responses to climate change , there are several studies which have demonstrated that responses can be heritable. These include examples of crop plants responding to environmental changes in situ and invasive plants or domesticates encountering new climate regimes as they expand their distribution [63–67]. Differential tracking of contemporary climate change has been shown to be a determinant of the species composition of a community, since populations of species lacking a plastic or microevolutionary response are locally extirpated [68, 69]. These findings raise questions about the significance of phenological adaptation to past climate change, since the extent to which differential tracking of past climate has shaped a whole flora is at present unknown.
Our study has identified a footprint of past phenological change, set in motion over five million years ago, in a contemporary flora. A point that remains clear regardless of the relative roles of evolutionary changes and phenotypic plasticity is that the observed phenological shifts have had a major role in shaping the extant Cape angiosperm flora; lineages that underwent shifts in flowering phenology later speciated extensively. Our study also provides tentative evidence of differential tracking of climate change, with phenological shifts apparent in some clades but not others.
The comparable number of shifts - distributional and phenological - experienced by the Cape flora as a new climate regime was established is a noteworthy finding of our study. Table 1 shows that phenological shifts are apparent in some clades (Crotalarieae, Disa, Heliophila, Moraea, Muraltia, Oxalis, Pentaschistis, Phylica and Podalyrieae) but not others. We go beyond estimation of the frequency of shifts to estimate the scale of the contribution these shifts have made to the character of the present-day flora. While our confidence intervals are wide - 14-41% and 14-55% of species belonging to lineages that have experienced distributional and phenological shifts consistent with past climate change, respectively - these figures indicate that contrary to certain schools of thought [18, 20], adaptive changes consistent with past climate change have had a significant impact on the Cape flora on the timescale considered (Early-Mid Miocene to the present). Such adaptations may be close contenders to (often co-occurring) distributional shifts in their frequency and contribution to the modern flora. Our results, while suggestive of evolutionary changes, do not allow us to rule out a major role for plasticity in the phenotypic adaptations that we observe. Furthermore, flowering phenology is just one of many important niche parameters of Cape plants, most of which are presumably highly conserved, otherwise there would not be large clades restricted to the Cape (Cape clades), and the Cape Floristic Region as we know it would not exist. Nonetheless, the patterns uncovered demonstrate that both adaptive and distributional shifts consistent with past climatic change have had strong impacts on the assembly of this biodiversity hotspot, since lineages that underwent these changes went on to contribute a high proportion of its current species richness.
While most of the observed shifts in distribution and flowering patterns are consistent with predictions based on past climate change, alternative explanations must be considered. First, we consider the possibility that reconstructed shifts in flowering patterns result from changes in distribution within the Cape, with a consequent phenological response to regions with differing climates, rather than climatic change itself. This scenario can be rejected because reconstructed shifts in flowering pattern (long-duration summer flowering to shorter-duration spring flowering) work in the opposite direction to predictions on the basis of reconstructed west to east shifts in geography.
Second, we consider the possibility that reconstructed shifts are an artefact of regional differences in species diversity within the Cape; the west is known to harbour higher species diversity than the east . Were the flowering duration or season of Cape species distributed at random with respect to phylogeny, we might expect the predominant flowering states observed in the west (short-duration spring flowering) to be more frequently reconstructed at the base of Cape clades than those observed in the east (longer-duration summer flowering), purely as a result of them being the most frequent in the dataset. This second scenario can be rejected, since observed shifts in flowering phenology run in the opposite direction to those predicted as a result of the proposed diversity difference artefact.
Third, we consider the possibility that the patterns we observe have arisen by chance. Our character randomizations of flowering midpoint across trees showing the basal-most (Podalyrieae) and distal-most (Disa) shifts in flowering midpoint showed that the probability of patterns consistent with past climatic change occurring by chance are P = 0.00 and P = 0.1 respectively. If we take the higher of these two probabilities as representative of the maximal probability of shifts consistent with past climatic change occurring by chance in any particular Cape clade, and consider the five out of 18 Cape clades for which shifts consistent with past climatic change arose in the real data, we can reject the hypothesis that the observed shifts occurred by chance (P = 0.0218).
It is appealing to link environmental changes and shifts in phylogenies through estimations of the timescales on which they occurred. In the case of the seasonal aridification trend in the Cape flora, problems with such an approach arise. The most significant of these is the absence of data concerning the precise timing, frequency and nature of the aridification trends; much more palaeobotanical and geological evidence is needed if we are to narrow down their timing reliably within the broad bounds of the Neogene. Notwithstanding this caveat, in all sampled Cape floral clades for which date estimates are available, confidence intervals for the timing of distributional and phenological shifts strongly overlap with temporal bounds of the aridification event (Table 1).
Since our phylogenies only sample living species, we are unable to speculate on any responses of non-surviving lineages of the early flora prior to their extinction. Clearly however, our conclusions regarding the type of response of lineages that survived past climatic change are unaffected by the type of response (or lack of response) of lineages that became extinct. Further to complete extinction, two patterns of local (regional) extinction may be envisaged. The first is extinction within the CFR with the lineage surviving as a "non-Cape relative" outside the CFR; we refer to this as 'Cape departure'. The second is relatively rapid back-and-forth shifting of distribution, leaving no significant distributional change between start and finish; we refer to this as 'back-and-forth shifting'.
Cape departure seems unlikely to present a major distortion of the broad-scale pattern presented here; such cases are infrequent judging by the small number of complete departures from the CFR occurring within the monophyletic (or nearly so) groups of Cape species ('Cape floral clades'). Therefore compared with most continental floras, the CFR can be viewed as being close to a closed system. Under the back-and-forth shifting scenario, a lineage's temporary absence accompanied by a shift in phenology and reinvasion of the CFR could have been misinterpreted as in situ phenological change. However, we have to invoke local extinction in the Cape at the time that the lineage colonized the neighbouring region, followed by local extinction in that neighbouring region as the lineage re-colonized the Cape. While this is not impossible it should first be noted that virtually any interpretation of a phylogeny can be refuted if enough extinction events in the right place are hypothesised. The more extinction events that are needed, and the more specific the placement of the extinct species required, the less strong such counter-arguments may seem. Here we require a minimum of two for each Cape clade inferred to undergo a phenological shift, making 18 local extinction events in total. More importantly, this scenario still involves phenological shifts. The only difference is that the phenological shifts occur while the lineages are outside the Cape, rather than being in situ. This would be an important detail to note were evidence found in favour. However, our main conclusion - that a large proportion of Cape clades have undergone phenological shifts during a period of past climatic change - would remain unchanged.
The cases of phenological shifts that we document provide a highly conservative estimate of the likely frequency of niche shifts in the Cape flora overall; we investigated shifts in timing and duration of flowering only, and there are many other pathways by which plants may have survived summer aridification events. These include the evolution of an annual life form, and changes in leaf longevity, sclerophylly, leaf size, root depth and root storage, all of which are interpreted by plant physiologists as mechanisms of drought resistance [70–74]. Improvements in a species' capability to survive a higher incidence of fire  through resprouting or reseeding may also have been important. While many of these species traits might be ideal for testing for adaptation to aridification between closely related species, few if any of them are likely to have been as important responses across the taxonomic breadth of the Cape flora as is flowering phenology.